To throw my own answer out there, I'd say that we should revise our definition of system guessing. Whilst there may be plenty of ambiguity between, say, a Pathfinder question and a D&D question, or between different editions of D&D, and that we should still generally ask for clarification from the OP as we've been doing as per our established policy, I feel that there are certain scenarios whereby we can definitively deduce, not guess, the system, and thus should not be considered system guessing, not subject to the same judgement.
I will point out that my answer is solely from a D&D perspective, since D&D 5e is the only game I really know about. But then, editions of D&D is also where we most commonly see this problem.
When the OP has linked to another Q&A with a system that matches our suspicions
Let's get the obvious one out of the way, the one that applies most directly to the question that I raised this meta question about. The OP made reference to another question, which was about D&D 5e and was tagged as such. It stands to reason that if the OP thought it relevant enough to link, there's a good chance they're asking about the same game, especially since they'd have recognised the rules being discussed in the other question.
It's not proof, because there's a chance that the OP is so unaware of the fact that there are other editions of D&D, and so unfamiliar with the rules of the edition that they are playing, that they could mistakenly think that a Q&A about a different edition is relevant to their question when it isn't, but I'd say the chances of that are so small that it's worth assuming it would never happen.
They reference exact text, maybe even with page numbers that match up with text
If the provide a quote, and the wording of the quote matches exactly with the corresponding text of the game/edition that we suspect, then I'd say that's good enough to say with certainty that we know what game they're talking about. I think it would be highly unlikely for two games, even two editions of the same game, to use the exact same text for a specific rule (although this is based on D&D, I must admit), so if we can confirm the exact quote, that should be good enough to say that we know the game/edition the OP is talking about.
I have seen in the past, not only an exact quote, but with a page number that matched up with the page number of the quote that matched up with the OP's quote, and yet, because they didn't explicitly say "D&D 5e" anywhere, was closed as unclear (one such example is this question, not with page numbers exactly, but a year of publication of the UA article in question, see version 1 of the edit history). Obviously this is ridiculous to claim that it's unclear when quotes and page numbers match up, so this is another case where we can safely say we know the game/edition without it being considered a guess.
There may be other criteria, but those are the two that stand out to me currently
There are likely other ways that we can safely deduce the game/edition from the content, but at the same time, we don't want to read into things too much because we think we recognise something. For me to list other ways of deducing a system alongside the two things I mention above, they would need to give me as much confidence and certainty as the things I list above such that I don't feel like I'm just guessing.
We shouldn't get ahead of ourselves and guess whilst hiding behind the excuse of "but it's obvious"
I noticed today a Pathfinder question had been asked about an Oath of Vengeance paladin. It was tagged as a pathfinder-1e question, but had it not have been, I might have thought "Oath of Vengeance? That's clearly D&D 5e because only D&D 5e has that exact subclass, right?", so we do still need to be careful and check our assumptions.
Making reference to spells and class features and other rules that we may recognise as being specific to D&D, or even a specific version of D&D, should not be considered enough to guess the system unless we know for a fact (as in, do our due diligence to actually research it before making an assumption) that it really is exclusive to the game/edition we suspect. Doing otherwise is guessing.
A final note about teaching the user
I've been fairly sold on the "making the user tell us their game/edition teaches them about the site and the importance of tagging" argument for a while now, but I have been swayed on that position by Cubic's answer (I'll quote the relevant parts here):
[...] new users, who'll simply not know the exact format their question should be in to be acceptable and who'd be much better served by simply being shown what an acceptable submission looks like rather than going through repeated stages of revisions until it is being fixed. Some might argue that this process has some effect of training new users in how to use this site correctly. I'd argue it is not really our place to play pedagogue, and anyone who's sticking around long enough for it to be even worth considering "training" them will pick up on the rules and norms of this community on their own over time.
I'm pretty sold on this argument. A user may come and ask one question, maybe not even ever come back to look at it (or forget how to log in again, so even if they did "look" at it again, they can't interact with it again), and therefore will not care if their question is tagged properly and don't have the interest in learning how our tagging system works because they're not planning on sticking around and becoming part of our community. Those that are interesting in sticking around, they will likely pick it up as they integrate themselves.
It might still be worth leaving a comment after adding such a system tag (again, only if we're truly sure, see above) telling the user that they really should be adding system tags themselves, in which case either they won't care because they'll never come back, or they will care and will learn to do so in their next question (or maybe after a few more reminders if they pick up more slowly than that).
Some might argue that a question asked by a user who does a hit-and-run question without ever coming back has no value, but since others may come by with a similar question in the future, I'd say that such questions do still have value if we can determine what system the OP was asking about and make it explicit in the tags for the benefit of future users.
In short, overall I agree with putting questions on hold and getting the OP to confirm the system if there is any ambiguity. In cases like this, however, where it is a certainty but is not explicitly mentioned by the OP, I think it makes no sense to follow this protocol and we should instead lead by example, adding the appropriate tag and informing the new user that they should do the same in future, assuming they want a future here.